Industrial Case Study: Moving to Commercial Off-the-Shelf and Open-Source Software Usage in Telecommunications
- 12/22/2011
- Setting the stage
- Options, recommendations, and reactions
- Outcomes and lessons learned
- Summary
- Web resources
Options, recommendations, and reactions
You believe that you have everything you need to ask for more money to cover the tasks associated with the use of COTS and open-source software on the development project. You have a rationale for using packaged software, a risk mitigation plan, a package selection list, plus a rationale for your choices, a task list for the work involved, and an associated cost estimate.
You are absolutely shocked when comments are received on the plan. Reviewers from the engineering department sliced your package selection rationale apart. They provided a more technical analysis that justified the use of custom rather than COTS and open-source packages for everything except the system software. They argue in detail that the risks associated with using someone else’s software are unacceptable because of performance and security issues. They further suggest that COTS and open-source software cannot be segmented easily to run on parallel processors even with their new scheduling and dispatching software. They point to vendor websites and blogs to provide evidence of security vulnerabilities and breaches that occurred in most of the packages. Based on the input, there is no doubt in anyone’s mind that they want to develop the software in-house. When you talk with the team to find out why, you get a big surprise. Several teammates tell you the reason is that the engineering department has a declining workload. They believe the department is using the new development project to justify the jobs for hundreds of people who will be laid off if packaged software is used for the new system.
You are caught in a quandary. You believe that going with COTS and open-source software is the right decision, but you do not want to be responsible for layoffs. The team lead understands your dilemma and suggests that the decision be elevated to senior management. According to the grapevine, engineering department leadership is already working the issue and is using their influence to squash the use of COTS and open-source software in the new system.
You decide the best way to compare the custom options against the COTS options is to look at the time and effort required to complete them. Then, if management wants to ignore the facts, that is their decision. You develop Table 4-5 to make the results easy to comprehend. For simplicity, you include only the software-related development costs across the three-year project schedule. To develop a credible development estimate, you use the parametric COCOMO II cost model4 using the standard estimating practices endorsed by the firm. Size estimates are those that engineering developed while performing a study to determine how many processors would be needed to accommodate peak loading on the new switching system. Results have been scaled proportionately for support tasks like product and project management using similar percentages.
Table 4-5 Custom vs. COTS and open-source software development comparison.
Task |
Custom Software Development |
COTS and Open-Source Package Usage |
||
Schedule (months)a |
Effort (in millions)b |
Schedule (months)c |
Effort (in millions)d |
|
Systems engineering |
6 |
$ 25e |
6 |
$ 30 |
Project management |
Level of effort |
75 |
Level of effort |
45 |
Product support |
Level of effort |
55 |
Level of effort |
30 |
Software engineering |
66 |
370 |
24 |
201 |
System integration and test |
12 |
55 |
12 |
30 |
Systems test |
3 |
10 |
3 |
10 |
Deployment |
3 |
5 |
3 |
5 |
TOTALS |
72 |
$595 |
36 |
$351 |
a Tasks will be done in parallel to achieve delivery in 3 years. bThe shortest feasible schedule according to COCOMO II model as calibrated is 6 years. cEffort was computed using a nominal profile for telecommunications and a size base of 4 MSLOC (million source lines of code). Cost/staff-year of effort is assumed $200,000. dNumbers were taken from Table 4-1 in this chapter and scaled to include software costs only. eCOCOMO II model estimates the cost is less than the estimates submitted by the engineering staff. |